MongoDB Extended JSON Converter

About this tool

MongoDB stores documents as BSON, which has types plain JSON does not: ObjectId, 64-bit integers, Decimal128, binary, regular expressions, timestamps. Extended JSON is the text encoding that keeps those types alive, by wrapping each value in a $-prefixed marker — {"$oid": "..."}, {"$date": "..."}, {"$numberLong": "42"}. It is exactly what mongoexport, driver EJSON helpers, aggregation playgrounds, and change-stream logs emit, and it is noisy to read or paste into an app that just wants ordinary JSON.

Use Extended JSON → plain JSON when you copied a document out of a mongoexport dump, a log line, or a driver response and want a readable object. Use Plain JSON → Extended JSON when you are preparing a fixture, a mongoimport file, or a test document that must carry exact types. Auto-detect unwraps input that already contains a recognised $-wrapper and wraps anything else.

The dialect control also makes this a canonical↔relaxed normaliser: pick Plain JSON → Extended JSON with Extended JSON already in the box, and the document is re-emitted in whichever dialect you chose. Relaxed keeps bare numbers and ISO-8601 dates for readability; canonical wraps every number and date so the exact BSON type survives a round trip.

Conversion is deterministic and entirely local: no MongoDB driver, no connection string, no network request. It only reshapes the JSON you paste.

Worked examples

Extended JSON input:

{"_id":{"$oid":"507f1f77bcf86cd799439011"},"created":{"$date":{"$numberLong":"1721485800000"}},"views":{"$numberLong":"42"},"name":"Ada"}

Plain JSON output:

{
  "_id": "507f1f77bcf86cd799439011",
  "created": "2024-07-20T14:30:00Z",
  "views": 42,
  "name": "Ada"
}

Plain JSON input, converted to the canonical dialect with Detect ObjectIds and dates in strings turned on:

{"_id":"507f1f77bcf86cd799439011","created":"2024-07-20T14:30:00Z","views":42,"score":1.5}

Canonical Extended JSON output:

{
  "_id": { "$oid": "507f1f77bcf86cd799439011" },
  "created": { "$date": { "$numberLong": "1721485800000" } },
  "views": { "$numberInt": "42" },
  "score": { "$numberDouble": "1.5" }
}

Limits and edge cases

• Document key order is preserved. BSON is an ordered format, so re-sorting keys can change meaning (index specs, command documents, $-operator order).
• Legacy spellings are accepted on input and normalised on output: {"$date": 1721485800000}, {"$binary": "...", "$type": "00"}, and {"$regex": "...", "$options": "i"} all parse.
• DBRef is not a wrapper. {"$ref": ..., "$id": ..., "$db": ...} is an ordinary subdocument in BSON, so it round-trips as one; its $id is still converted.
• Infinity, -Infinity, and NaN always unwrap to strings, because JSON has no literal for them. There is no lossless numeric alternative to offer.
• Very large numbers. JSON parsers in JavaScript lose precision above 2^53 − 1. Turn on Keep 64-bit and decimal values as strings to unwrap $numberLong and $numberDecimal as strings instead of numbers.
• Dates outside years 1970–9999 keep the canonical {"$date": {"$numberLong": ...}} form in relaxed mode, matching the specification rather than emitting an out-of-range ISO string.
• Binary values pass through as base64 plus their subtype; the tool does not decode binary payloads or reinterpret legacy UUID subtypes.
• Input must be a single JSON value. Newline-delimited JSON dumps need splitting into individual documents first.

FAQ